A prospective study was conducted among different intra and extra-hospital populations of French Guiana to evaluate the performance of saliva testing compared to nasopharyngeal swabs. Persons aged 3 years and older with mild symptoms suggestive of COVID-19 and asymptomatic persons with a testing indication were prospectively enrolled. Nasopharyngeal and salivary samples were stored at 4°C before analysis. Both samples were analyzed with the same Real-time PCR amplification of E gene, N gene, and RdRp gene. Between July 22th and October 28th, 1159 persons were included, of which 1028 were analyzed. When only considering as positives those with 2 target genes with Ct values <35, the sensitivity of RT-PCR on saliva samples was 100% relative to nasopharyngeal samples. Specificity positive and negative predictive values were above 90%. Across a variety of cultures and socioeconomic conditions, saliva tests were generally much preferred to nasopharyngeal tests and persons seemed largely confident that they could self-sample. For positive patients defined as those with the amplification of 2 specific target genes with Ct values below 35, the sensitivity and specificity of RT-PCR on saliva samples was similar to nasopharyngeal samples despite the broad range of challenging circumstances in a tropical environment.